This thesis introduces new image compression algorithms, their related architectures and data transforms architectures. The proposed architectures consider the current hardware architectures concerns, such as power consumption, hardware usage, memory requirement, computation time and output accuracy. These concerns and problems are crucial in multidimensional image and video processing applications. This research is divided into three image and video processing related topics: low complexity non-transform-based image compression algorithms and their architectures, architectures for multidimensional Discrete Cosine Transform (DCT); and architectures for multidimensional Discrete Wavelet Transform (DWT). The proposed architectures are parameterised in terms of wordlength, pipelining and input data size. Taking such parameterisation into account, efficient non-transform based and low complexity image compression algorithms for better rate distortion performance are proposed. The proposed algorithms are based on the Adaptive Quantisation Coding (AQC) algorithm, and they achieve a controllable output bit rate and accuracy by considering the intensity variation of each image block. Their high speed, low hardware usage and low power consumption architectures are also introduced and implemented on Xilinx devices. Furthermore, efficient hardware architectures for multidimensional DCT based on the 1-D DCT Radix-2 and 3-D DCT Vector Radix (3-D DCT VR) fast algorithms have been proposed. These architectures attain fast and accurate 3-D DCT computation and provide high processing speed and power consumption reduction. In addition, this research also introduces two low hardware usage 3-D DCT VR architectures. Such architectures perform the computation of butterfly and post addition stages without using block memory for data transposition, which in turn reduces the hardware usage and improves the performance of the proposed architectures. Moreover, parallel and multiplierless lifting-based architectures for the 1-D, 2-D and 3-D Cohen-Daubechies-Feauveau 9/7 (CDF 9/7) DWT computation are also introduced. The presented architectures represent an efficient multiplierless and low memory requirement CDF 9/7 DWT computation scheme using the separable approach. Furthermore, the proposed architectures have been implemented and tested using Xilinx FPGA devices. The evaluation results have revealed that a speed of up to 315 MHz can be achieved in the proposed AQC-based architectures. Further, a speed of up to 330 MHz and low utilisation rate of 722 to 1235 can be achieved in the proposed 3-D DCT VR architectures. In addition, in the proposed 3-D DWT architecture, the computation time of 3-D DWT for data size of 144×176×8-pixel is less than 0.33 ms. Also, a power consumption of 102 mW at 50 MHz clock frequency using 256×256-pixel frame size is achieved. The accuracy tests for all architectures have revealed that a PSNR of infinite can be attained.

This thesis summarises research into detecting moving objects against moving backgrounds using a pixel-wise classification. Initial experimentation with static background systems is presented along with efforts to extend such techniques to moving background situations. The results of this approach are studied and the deficiencies identified. A new system using the position is then presented after theoretical investigation of the problem which uses the rank of the displaced frame difference (DFD). This yields superior results to the DFD alone. Further work for consolidating the resultant object masks and tracking them temporally using a CONDENSATION based algorithm is also presented.

This thesis investigates the potential of computer graphics in providing for a graphics driven specification system that gives sufficient structure and content to form the simulation model itself. The nature of discrete event simulation modelling, the diagramming method of activity cycle diagrams which underpinned this research, the three phase simulation model structure, and the trend of visual simulation modelling are discussed as the basis for the research. Some current existing simulation languages and packages are reviewed, which gives insight into the essential features of an ideal computer simulation environment. The basic research method adopted was to build systems that exemplified the state of thinking at the time. The purpose of this method was to enable ideas to be developed, discarded and enhanced, and for new ideas to emerge. The research has undergone a series of application developments on the Apple Macintosh to examine the advantages and limitations of such systems. The first system developed during the research, MacACD, provides the basis for proposals concerning the enhancement of the ACD diagramming method in a computer-aided environment. However, MacACD demonstrated the limitations of an ACD interface and the need for a more flexible specification system. HyperSim, a simulation system developed using HyperCard, has all the power of interconnectivity demonstrated as a need by MacACD, but has severe limitations both in terms of security of system development, and an inability to provide a running model directly due to lack of speed. However, the power of an icon-based interconnected textual and diagrammatic based system were demonstrated by the construction of this system during this research, and led to the development of the final system described in this thesis : MacGraSE. The development of this system during this research incorporates many innovations. The main input device is a picture representing the problem, including a background display. This system allows for dynamic icon based visual model running, as well as code generation for complete model embellishments, interactive report writing, and representational graphics outputs.

Personal computing is continuously moving away from traditional input using mouse and keyboard, as new input technologies emerge. Recently, natural user interfaces (NUI) have led to interactive systems that are inspired by our physical interactions in the real-world, and focus on enabling dexterous freehand input in 2D or 3D. Another recent trend is Augmented Reality (AR), which follows a similar goal to further reduce the gap between the real and the virtual, but predominately focuses on output, by overlaying virtual information onto a tracked real-world 3D scene. Whilst AR and NUI technologies have been developed for both immersive 3D output as well as seamless 3D input, these have mostly been looked at separately. NUI focuses on sensing the user and enabling new forms of input; AR traditionally focuses on capturing the environment around us and enabling new forms of output that are registered to the real world. The output of NUI systems is mainly presented on a 2D display, while the input technologies for AR experiences, such as data gloves and body-worn motion trackers are often uncomfortable and restricting when interacting in the real world. NUI and AR can be seen as very complimentary, and bringing these two fields together can lead to new user experiences that radically change the way we interact with our everyday environments. The aim of this thesis is to enable real-time, low latency, dexterous input and immersive output without heavily instrumenting the user. The main challenge is to retain and to meaningfully combine the positive qualities that are attributed to both NUI and AR systems. I review work in the intersecting research fields of AR and NUI, and explore freehand 3D interactions with varying degrees of expressiveness, directness and mobility in various physical settings. There a number of technical challenges that arise when designing a mixed NUI/AR system, which I will address is this work: What can we capture, and how? How do we represent the real in the virtual? And how do we physically couple input and output? This is achieved by designing new systems, algorithms, and user experiences that explore the combination of AR and NUI.

High-performance computer graphics is becoming more demanding due to the increasing use of large 3D scenes. The visibility determination is one of the fundamental computational problems in computer graphics. In the past couple of decades many researchers developed a variety of visibility determination algorithms such as visibility scanline algorithms, distributed ray tracing and output sensitive visibility algorithms to determine the hidden parts of a complex 3D scene. If hidden surface determination can be solved in a cost-effective way, the overall system performance will be improved dramatically. If not, it may become a bottleneck in terms of performance. One of the main aims of this dissertation is to give a solution to the rendering problem using a scanline-based distributed system to increase rendering power. The features of this system are to use the processing power of idle processors over the network to balance workload and reduce communication overhead. The system does not use any extra resources (memory or graphics processing unit (GPU) chips) to handle large data sets. Using a network of workstations, the proposed system could be more powerful than a GPU and binary swap compositing for high-performance graphics. Visibility computation, clipping and transformation are the basic functions required for rendering a static scene. The visibility computations exhibit a quadratic growth rate. If the number of objects or depth of the scene increases, the computational requirement can easily exceed to the performance of a single processor. Polygon clipping has a growth rate of N log N in the worst case where N is the total number of edges in the scene. The time requirement for transformations grows linearly with the input size. This dissertation presents the requirement of real-time rendering of 3D scenes and a few approaches, including parallel techniques, to speed up the rendering process. All existing visibility scanline algorithms – including the new one proposed within this thesis – are presented and discussed with respect to their potential for use in the context of distributed systems for the visualisation of large data sets. The proposed scanline algorithm has many advantages, including: using real values obtained by the intersection of a 3D scene with the plane of the scanline as input; and running time that does not depend on the resolution over the Z-tree scanline algorithm. Z-tree, Warnock, Priority Queue and proposed algorithm have been implemented in C/C++ and Java program to show how they compute visibility in a 2D scene. Design issues of all scanline algorithms and tuning cache parameters (without knowing them) to minimise cache misses and data movement among multiple levels of caches have been presented in the light of cache-oblivious algorithms. The cache-oblivious scanline algorithms have been optimised for use within a distributed system. The system performance can be increased by the careful consideration of locality of cache reference, instruction-level parallelism, branch prediction and memory hierarchies found in modern computers. A new scanline-based approach for the distribution of workload among several servers and gathering processed data for the visualisation of large data sets is proposed in this thesis. Based on this approach, a prototype scanline-based distributed system has been developed and tested over three servers giving input of line segments from 100 to 3,000 over 500 × 500 and 3,000 × 3,000 pixels visualisation area; and is demonstrated with its system operation and compression techniques in order to handle large data sets. While considering the efficiency of the algorithms asymptotic analysis is often used. However, asymptotic analysis cannot take into consideration the constant factors that could be different in different environments. Therefore, the author developed a portable test bed for the comparative evaluation of the actual performance of the algorithms for the application of geometric algorithms (line-clipping and few scanline algorithms). For this thesis, experimental evaluation of geometric algorithms is important to design, implement and test from different perspectives and then choose the best one for high-performance computer graphics system. Through experimental evaluation of line-clipping algorithm, the author found that clock cycle counting is more accurate than using elapsed-time functions provided by system software. The author also implemented, measured and evaluated running time of Z-tree, Warnock, Priority Queue and new Sorting and Ranking (SR) scanline algorithms for 1 to 5,000 line segments with 500 × 500 pixels, for 1 to 25,000 line segments with 3,000 × 3,000 pixels and for 1 to 250,000 line segments with 100,000 × 100,000 pixels visualisation area. An evaluation technique based on best sustained performance is proposed in the thesis as well. To conclude, this thesis presents new tools and techniques, including a scanline-based distributed systemand its operation for the visualisation of very large data sets without any additional cost.

The main goal of this thesis is to review and offer robust and efficient algorithms for the detection (or the segmentation) of foreground objects in indoor and outdoor scenes using colour image sequences captured by a stationary camera. For this purpose, the block diagram of a simple vision system is offered in Chapter 2. First this block diagram gives the idea of a precise order of blocks and their tasks, which should be performed to detect moving foreground objects. Second, a check mark (✓) on the top right corner of a block indicates that this thesis contains a review of the most recent algorithms and/or some relevant research about it. In many computer vision applications, segmenting and extraction of moving objects in video sequences is an essential task. Background subtraction has been widely used for this purpose as the first step. In this work, a review of the efficiency of a number of important background subtraction and modelling algorithms, along with their major features, are presented. In addition, two background approaches are offered. The first approach is a Pixel-based technique whereas the second one works at object level. For each approach, three algorithms are presented. They are called Selective Update Using Non-Foreground Pixels of the Input Image , Selective Update Using Temporal Averaging and Selective Update Using Temporal Median , respectively in this thesis. The first approach has some deficiencies, which makes it incapable to produce a correct dynamic background. Three methods of the second approach use an invariant colour filter and a suitable motion tracking technique, which selectively exclude foreground objects (or blobs) from the background frames. The difference between the three algorithms of the second approach is in updating process of the background pixels. It is shown that the Selective Update Using Temporal Median method produces the correct background image for each input frame. Representing foreground regions using their boundaries is also an important task. Thus, an appropriate RLE contour tracing algorithm has been implemented for this purpose. However, after the thresholding process, the boundaries of foreground regions often have jagged appearances. Thus, foreground regions may not correctly be recognised reliably due to their corrupted boundaries. A very efficient boundary smoothing method based on the RLE data is proposed in Chapter 7. It just smoothes the external and internal boundaries of foreground objects and does not distort the silhouettes of foreground objects. As a result, it is very fast and does not blur the image. Finally, the goal of this thesis has been presenting simple, practical and efficient algorithms with little constraints which can run in real time.

Link Analysis (LA) is a visual data analysis technique originally developed for analysing crime related data. The technique enables the user to gain a better understanding of the connections between objects of interest in the problem domain by displaying the connections in a form of network diagram referred to as a link chart. A link chart is often altered during its lifetime, as part of the exploratory nature of the knowledge discovery process, to reflect new information, and to increase the level of comprehensibility. To provide the necessary flexibility for accessing and manipulating large volumes of data the data collected in an investigation is often stored in a database. Permitting link charts to be constructed from this data is of great value, as LA is not adequately supported by the database query systems currently available. This is because the record-based data models they often use are inappropriate for modelling connections between objects, the query languages they typically provide can only retrieve connections between objects if the way connections may be derived are specified in the query, and their data visualisation facilities generally do not allow the results of a number of queries to be integrated and edited. This thesis concerns the Exploratory Database View Constructor (EDVC), an experimental visual database interface for supporting LA. The results obtained from a user evaluation of EDVC indicate that the system may be used by individuals with no experience of interacting directly with a database management system (DBMS). This is a consequence of the style of interaction supported, which allows the data stored to be browsed without having to possess explicit knowledge of the database schema. Such knowledge is typically a prerequisite for using a database query language and can prevent productive use of such a language by an inexperienced DBMS user.

High dynamic range (HDR) technology (capture and display) call offer high levels of immersion through a dynamic range that meets and exceeds that of the Human visual system (HVS). This increase in immersion comes at the cost of higher bit-depth, bandwidth and memory requirements, which are significantly higher than those of conventional Low dynamic range (LDR) content. The challenge is thus to develop a coding solution to efficiently compress HDR images and video into a manageable bitrate without compromising perceptual quality. Over the past century, a large number of psycho-visual experiments have been carried out by psychologists and physiologists with the goal of understanding how the HVS works. One of the human vision phenomena concerns reduced sensitivity to patterns of low and high spatial-frequencies. This phenomenon is parametrized by the contrast sensitivity function (CSF). In this thesis, proper luminance and chrominance CSFs have been employed, in conjunction with an optimised wavelet 1mb-band weighting method. Experimental results indicate that the proposed method ontperforms previous approaches and operates in accordance with the characteristics of the HVS, when tested objectively using a HDR Visible Difference Predictor (VDP), and subjective evaluation. The HVS shows non-linear sensitivity to the distortion introduced by lossy image and video coding. Two psycho-visual experiments were performed using of a high dynamic range display, in order to determine the potential differences between LDR and HDR edge masking (EM) and luminance masking (LM) effects. The EM experimental results indicate that the visibility threshold is higher for the case of HDR content than for LDR, especially on the dark background side of an edge. The LM experimental results suggest that the HDR visibility threshold is higher compared to that of SDR for both dark and bright luminance backgrounds. A novel perception-based quantization method that exploits luminance masking in the HVS in order to enhance the performance of the High Efficiency Video Coding (HEVC) standard for the case of HDR video content has been proposed in this thesis . The proposed method has been integrated into the reference codec considered for the HEVC range extensions and its performance was assessed by measuring the bitrate reduction against the codec without perceptual quantization. The results indicate that the proposed method achieves significant bitrate savings, up to 42.2%, compared to HEVC at the same objective quality (based on HDR-VDP- 2) and subjective evaluation.

Video compression is necessary for effective coding of video data so that the data can be stored or transmitted more efficiently. In video compression, the more redundant data are discarded, the higher compression ratios will be achievable. This causes the contents of a compressed bitstream to be highly dependent on each other. In video communications, the compressed bitstream is ,subject to losses and errors due to the nature of the transmission medium. Since the contents of the compressed bitstream are highly dependent on each other, when a loss or an error occurs this leads to propagation of the error, which causes deterioration of the decoded video quality. Error resilience plays an important role / in decreasing the quality degradation caused by losses arid errors. Error resilience methods can either take place in the encoder side as a coding technique which decreases the effects of errors on the coded bitstream or in the decoder side as a technique which conceals the detected errors or losses. Error concealment which takes place in decoder side and redundant slice coding which takes place in encoder side are investigated throughout the thesis. The first part of the thesis investigates efficient error concealment techniques for Scalable Video Coding (SVC). These include the utilisation of higher Temporal Level picture motion information and the utilisation of "Bridge Pictures" which will be described in later chapters, for error concealment. The second part of the thesis investigates redundant slice coding for SVc. Single Block per Macroblock and Zero Residual redundant slice coding schemes are proposed and tested in this part of the thesis. In addition to these, an adaptive redundant slice allocation scheme is also proposed and tested. The last part of the thesis investigates error resilient coding techniques for multi-view 3D video. Multi-view 3D video compression is achieved using the SVC CoDec by coding one of the views as the Base Layer and the other views as the Enhancement Layers utilising the adaptive inter-layer prediction mechanism of the SVc.

The advances in multi-core architecture for general-purpose computing in the past decade have tremendously increased the available raw computing power. The two major architectures are the central processing unit (CPU) and the graphics processing unit (CPU). CPUs have been developed recently as general purpose processors. The present work is focused on the performance of unstructured CFD solvers on the CPU. For this purpose an explicit and implicit solvers were developed. The explicit solver for the CPU and the multicore CPU were generated using the OPlus 2 library. This was achieved by implementing minimal extensions to the sequential code. The explicit solver achieved a speedup of an order of magnitude on the CPU, compared to the multi-core CPU code. For the explicit solver the CPU is a cost effective option compared to the CPU. On the other hand, the implicit solver using the Jacobi linear solver was implemented in two variants. The first using the OPlus 2 library and the second using NVIDIA library. The manufacturer library performed better than the OPlus 2 implementation. This was due to the inefficient implementation of the OPlus 2 version. The NVIDIA library gave a speedup of 27x compared to the sequential version. Hence, for the implicit solver the CPU might not be a viable option.